Quick start
- From the dashboard choose New matrix, pick a type (start with Named segments if you're not sure) and choose Continue.
- Name it, pick your transactions (or upload them next) and keep the window at the last 3 years.
- Watch the cells fill in as everyone is scored.
- Click any cell, segment or profile for View records: the IDs behind it, ready to export as CSV and match back in your CRM or mailing tool.
- Change how scores are cut with Scoring, or rename segments and move cells with Edit segments.
What RFM is
RFM describes everyone you serve with three measures:
- Recency — how many days since their last gift or purchase.
- Frequency — how many gifts or purchases they made in the window.
- Monetary — their net amount in the window.
Each measure becomes a score from 1 to 5, where 5 is always best: the most recent, the most frequent, the most valuable. Someone who gave or bought last week, often and generously is a 5-5-5; someone last seen years ago, once, for a small amount, is a 1-1-1. Each matrix type puts these scores side by side in a different way.
The matrix types
Every type scores the same people the same way; they differ in which scores they show. Build as many as you like from one upload — each is its own matrix with its own page.
Named segments
- Champions930 · 56% of giving or spend
- Loyal donors or customers740 · 18% of giving or spend
- Potential loyalists1,510 · 8% of giving or spend
- Can't lose them180 · 6% of giving or spend
- At risk820 · 5% of giving or spend
The R × FM grid's 25 cells are grouped into named segments, each with what it means and a suggested next step, worded for your organization type. The page shows the grid colored by segment, a table of segments with their people, share of people, value and share of value (largest value first), and, when your data reaches back far enough, movement since a year before.
The standard 11 segments and the cells they start with (R is recency, FM is frequency + monetary):
| Recency | FM 1 | FM 2 | FM 3 | FM 4 | FM 5 |
|---|---|---|---|---|---|
| R 5 | New | Potential loyalists | Potential loyalists | Champions | Champions |
| R 4 | Promising | Potential loyalists | Potential loyalists | Loyal | Champions |
| R 3 | About to sleep | About to sleep | Need attention | Loyal | Loyal |
| R 2 | Hibernating | Hibernating | At risk | At risk | Can't lose them |
| R 1 | Lost | Lost | At risk | At risk | Can't lose them |
- Champions
- Recent, frequent and the most valuable. Thank them personally and invite them closer.
- Loyal donors / Loyal customers
- Give or buy often and well, though not the very latest. Keep them close and ask for more.
- Potential loyalists
- Recent, with a few transactions of moderate size. Nurture them into regulars.
- New donors / New customers
- Their first transaction was very recent. Welcome them quickly; the second one is the one that counts.
- Promising
- Fairly recent, but small and infrequent so far. Build the relationship.
- Need attention
- Middle of the road on everything. Re-engage before they slip.
- About to sleep
- Small and fading. A light-touch reminder; don't spend much here.
- At risk
- Used to give or buy often and well, but haven't for a while. Win them back now.
- Can't lose them
- Among your most valuable, gone quiet. Reach out personally.
- Hibernating
- Last transaction long ago, small and infrequent. Low-cost outreach only.
- Lost
- Lowest on everything. Try one appeal or promotion to win them back, then stop paying to reach them.
You can rename segments, rewrite what they mean and their next step, and move cells between them. See editing segments.
R × FM grid
The same 5 × 5 grid without names: recency (R5 at the top) against frequency + monetary (FM1 to FM5 across). Each cell shows its people, its share of people and its share of value, and is shaded darker the more people it holds. Use it to see the whole file at a glance and pick the cells to act on yourself. See reading the grid.
Full RFM
All three scores kept separate: 125 profiles from 5-5-5 down to 1-1-1. The page shows:
- Five grids, one per recency score from R5 to R1, each frequency (F5 at the top) against monetary (M1 to M5 across), with the number of people in each profile. Click a cell for its records.
- The composite score, R + F + M, as a bar chart from 15 down to 3. See the composite score.
- All profiles, a table of every profile with people in it. See profiles.
RF, FM and RM grids
Two-score grids for when one measure doesn't fit your model. Each is a 5 × 5 grid like the R × FM grid.
- RF grid — recency × frequency
- Leaves amount out. For memberships, subscriptions, content, or goods at standard prices, where amounts barely vary.
- FM grid — frequency × monetary
- Leaves recency out. For annual renewals and seasonal or project-based giving or buying, where repeat volume matters more than the last date.
- RM grid — recency × monetary
- Leaves frequency out. For high-value, infrequent transactions: major gifts, furniture, appliances, B2B contracts.
RFE (engagement)
Recency, frequency and engagement: activities such as events attended, volunteer shifts, logins or email opens, in place of amount. It needs an engagement file as well as your transactions — one row per activity with an ID that matches your transactions, a date and, if you have it, the activity type. You can choose one when you create the matrix or upload one later from the matrix page; until then everyone's engagement is 0 and scores 1.
- Engagement is the number of activities in the window.
- Recency uses whichever is later: the last counted transaction or the last activity.
- People with activities but no transactions are included, with frequency 0.
The grid shows recency against frequency + engagement (FE), combined the same way as FM.
True segments
Segments learned from your data by machine learning rather than score bands, with a “likely to move up” list. They're built differently from the matrices above; see True segments.
How scoring works
A build reads every counted transaction in the file, works out each person's measures, then turns each measure into a score with four breakpoints.
As of and the window
Every matrix is measured as of the date of the latest counted transaction in its file, not today, so an export from last month scores the same as one from this morning. Recency counts the days from each person's last transaction to that date.
Frequency, monetary and engagement count only the window: the last 1, 2, 3 or 5 years back from the as-of date, or all history. The default is 3 years. Recency always looks at the whole file, so someone whose last gift or purchase was before the window still appears, with low recency and frequency 0.
Quintiles (the default)
Each measure's breakpoints are the values at the 20th, 40th, 60th and 80th percentile of your file, so each score holds roughly a fifth of people. People with the same value always get the same score, so when many share a value the bands can be uneven. In most files more than half of people have exactly one transaction in the window, so they all get F1 and the higher frequency scores hold fewer people. That's expected.
Recency runs the other way: within the first breakpoint's days scores 5. For frequency, monetary and engagement, a value at or below the first breakpoint scores 1 and above the last scores 5.
Example: recency breakpoints
If your recency breakpoints are 60, 150, 320 and 700 days: last transaction within 60 days is R5, 61–150 days R4, 151–320 days R3, 321–700 days R2, over 700 days R1.
Fixed breakpoints
Choose Scoring on the matrix page to switch any measure to Fixed breakpoints and type the four numbers yourself: recency in months, frequency in transactions, monetary in dollars, engagement in activities. The boxes start from the breakpoints the last build used. Each number must be at least the one before it, or the panel says so and nothing is saved. Fixed breakpoints are useful when you want scores to mean the same thing every year, or to match an existing RFM code.
The panel shows what each score covers right now, and Save and score again rebuilds the matrix.
Combined scores: FM and FE
The R × FM grid and named segments combine frequency and monetary into one FM score: their average, with halves rounded up. F5 and M4 make FM5; F2 and M1 make FM2; F3 and M3 make FM3. RFE combines frequency and engagement into FE the same way.
Which transactions count
Which transaction types count follows your organization type:
| Organization type | Counts toward frequency and monetary | Never counts |
|---|---|---|
| Nonprofit and government | Money received: one-time and recurring gifts, pledge payments, in-kind gifts and memberships | Open pledges and planned gifts (they're promises, not money), purchases, subscriptions, tickets |
| Commercial | Purchases, subscriptions and tickets | Gifts, pledges, memberships |
Refunds always come off the amount, whatever their sign in the file, and are never counted as a transaction. A row of any type with a negative amount comes off too. Only rows with a positive amount count toward frequency and recency. Changing the organization type in Settings scores every matrix again.
While a matrix builds
A new matrix, a changed scale or window, edited segments, a new upload or a CRM refresh all start a build. The page shows three stages:
- Reading transactions by person — rows read so far.
- Setting the breakpoints — quintiles need everyone's numbers, so this waits for the whole file.
- Scoring — people scored so far. The cells and segment table fill in as each batch is saved; counts only grow.
The tag reads Still counting until it's done, and the page reloads by itself at the end. You can leave; building keeps going, and we email you when it's ready if you left.
If a build doesn't finish
The page says Scoring didn't finish with the reason and a Try again button. A build that was replaced by a newer one (because you saved a change while it ran) shows “Replaced by a newer build.” and the newer build carries on. If it keeps failing, upload the file again or email support@truerfm.com.
Reading the grid
Rows are the first score (recency, or frequency on the FM grid) from 5 at the top to 1 at the bottom; columns are the second score from 1 on the left to 5 on the right. Under each score's label is what it covers in this build, such as “Within 2 months” or “$250 – $1,000”; combined scores (FM, FE) show just the score.
Each cell shows its people, then % of people · % of value. The top right is recent and valuable; the bottom left is long gone and small. A cell with a large share of value but few people deserves the most personal attention; a large bottom-left share is where outreach costs the most for the least.
Click a cell to open its records. Cells with nobody in them can't be opened.
The numbers above the matrix
- Donors or customers scored
- Everyone with a counted transaction on or before the as-of date (or, for RFE, an activity).
- Giving or spend in the window
- Everyone's net amount in the window added together.
- Per donor or customer
- That total divided by the people scored.
- Top 20% hold
- The share of that total held by the fifth of people with the largest amounts in the window. Most files are concentrated: 60–80% is common.
Editing segments
On a named segments matrix, choose Edit segments.
- Rename any segment, and rewrite what it means and its suggested next step.
- Move cells: pick a segment with its radio button, then click cells of the grid to paint them into it. Every cell always belongs to exactly one segment.
- Save and score again rebuilds the matrix with the new map. Names must all be filled in and different from each other.
- Restore standard segments goes back to the 11 standard segments and map, replacing your names and cells.
You can't add or remove segments in the panel yet: rename one you don't need and paint its cells into another.
Movement since a year before
On a named segments matrix whose file reaches back more than a year before the as-of date, the page compares everyone with where they were a year earlier. Each person is scored as of one year before the as-of date, over the same length of window, using the same breakpoints as now — so a move means that person's own behavior changed, not that the rest of the file shifted.
- Same segment, Changed segment and New since then (people with no counted transaction a year before) are totaled at the top.
- The table's rows are where people were then, its columns where they are now; the last row is people who weren't there. Cells on the diagonal are people who stayed. Click any number for those people's records.
If the file starts less than a year before the as-of date, the page says movement was skipped and why. For changes between any two months, use Matrix history (Plus).
The composite score
On a full RFM matrix, each person's R + F + M, from 3 (1-1-1) to 15 (5-5-5). The chart shows how many people have each score and their share of value; View records opens everyone with that score. It's a quick way to make a selection such as “everyone 12 and above”.
Profiles
The full matrix's All profiles table lists every profile with people in it: profile, score, people, share of people, value and share of value. Sort it by profile (5-5-5 first), people, value or score. View records opens a profile's people.
How these numbers were made
Under every matrix, a short list states the assumptions behind it: the as-of date, the window and its dates, which transaction types counted, and each measure's breakpoints and whether they're quintiles or fixed. For RFE it also says which engagement file was used. Include them when you share the numbers.
Exporting
- Export CSV at the top of the page downloads the matrix's groups with their people and value: segments (with what they mean and their next step), profiles for the full matrix, or cells for the grids.
- View records › Export CSV downloads the IDs in any cell, segment, profile, score or movement group, with their measures and scores. See exporting records.
Questions
Why doesn't each score hold exactly a fifth of people?
Because people with the same value always get the same score. If 60% of your file made exactly one transaction in the window, they all score F1. Switch frequency to fixed breakpoints if you'd rather choose the bands yourself.
What happens when I change the window?
Frequency, monetary and engagement count a different stretch of time, the quintiles are worked out again, and everyone is scored again. Recency doesn't change.
Will two matrices from the same file agree?
Yes, if they have the same window and scales: every type scores people identically and only shows the scores differently.
How do I bring a matrix up to date?
Upload a new file and build from it, or, for a matrix built from your CRM, choose Refresh from CRM; with Plus, scheduled sync does it for you. See refreshing.